Kernel density estimation using randomized response models
- UNCG Author/Contributor (non-UNCG co-authors, if there are any, appear on document)
- Wenhao Shou (Creator)
- Institution
- The University of North Carolina at Greensboro (UNCG )
- Web Site: http://library.uncg.edu/
- Advisor
- Sat Gupta
Abstract: The randomized response technique (RRT) was first introduced to estimate prevalence of sensitive characteristics for binary response variables. Extensions to quantitative variables using additive and/or multiplicative scrambling were later explored for population parameter estimation, but estimation of population distribution estimation for sensitive variables remains underexplored. This dissertation investigates kernel density estimation (KDE) for sensitive variables using additive Randomized Response Technique (RRT) models, addressing the gap in direct distribution estimation in this field. It refines prior work on direct distribution for sensitive variables, particularly KDE under multiplicative RRT models, and explores KDE under additive RRT models. The research encompasses the application of KDE in the presence of auxiliary information and further study of KDE under optional RRT models. Simulation results show that the kernel density estimator using additive scrambling performs better and is more flexible in bandwidth selection compared to multiplicative scrambling. Additionally, the inclusion of auxiliary variables enhances the accuracy of sensitive variable distribution estimation. Introducing sensitivity level W into RRT models as an option proves beneficial under certain conditions for extreme values of W, or when noise levels are high. By combining the strengths of KDE and additive RRT models, this research seeks to contribute to the advancement of estimation techniques for sensitive variables and provide valuable insights into their distribution. The findings may enhance the understanding and application of survey sampling methodologies when dealing with sensitive and privacy-related information.
Kernel density estimation using randomized response models
PDF (Portable Document Format)
2735 KB
Created on 8/1/2024
Views: 39
Additional Information
- Publication
- Dissertation
- Language: English
- Date: 2024
- Keywords
- Auxiliary variable, Kernel density estimation, Mean integrated squared error, Randomized response models
- Subjects
- Surveys $x Statistical methods
- Variables (Mathematics)
- Sampling (Statistics)